An efficient Deep Spatio-Temporal Context Aware decision Network (DST-CAN) for Predictive Manoeuvre Planning

TL;DR

DST-CAN addresses predictive manoeuvre planning for autonomous vehicles by forecasting surrounding vehicles' future trajectories with a Memory Neuron Network and encoding past, present, and predicted futures into spatio-temporal context-aware grids. A CNN-based decision engine maps these grids and Probabilistic Occupancy Maps (POMs) to safe and efficient lateral and longitudinal maneuvers, trained via imitation learning on both human decisions and rule-based ground truth. Evaluated on the NGSIM US-101 and I-80 datasets, DST-CAN with a $3$-second horizon shows clear advantages over CS-LSTM, particularly in conflict scenarios, while also robustly handling near-collision cases. Key contributions include uncertainty-aware occupancy mapping, a data-pruning strategy for long-tail imitation data, and demonstrated improvements across horizons and decision-bias conditions, with potential extensions to incorporate a global drivable-area map and passenger comfort considerations.

Abstract

To ensure the safety and efficiency of its maneuvers, an Autonomous Vehicle (AV) should anticipate the future intentions of surrounding vehicles using its sensor information. If an AV can predict its surrounding vehicles' future trajectories, it can make safe and efficient manoeuvre decisions. In this paper, we present such a Deep Spatio-Temporal Context-Aware decision Network (DST-CAN) model for predictive manoeuvre planning of AVs. A memory neuron network is used to predict future trajectories of its surrounding vehicles. The driving environment's spatio-temporal information (past, present, and predicted future trajectories) are embedded into a context-aware grid. The proposed DST-CAN model employs these context-aware grids as inputs to a convolutional neural network to understand the spatial relationships between the vehicles and determine a safe and efficient manoeuvre decision. The DST-CAN model also uses information of human driving behavior on a highway. Performance evaluation of DST-CAN has been carried out using two publicly available NGSIM US-101 and I-80 datasets. Also, rule-based ground truth decisions have been compared with those generated by DST-CAN. The results clearly show that DST-CAN can make much better decisions with 3-sec of predicted trajectories of neighboring vehicles compared to currently existing methods that do not use this prediction.

Figures (8)

• Figure 1: A typical traffic situation with an AV in operation
• Figure 2: A schematic diagram for PMP using Deep Spatio Temporal Context-Aware decision Network (DST-CAN)
• Figure 3: MNN architecture for look-ahead trajectory prediction
• Figure 4: Context and Probabilistic Occupancy Map (POM) embedding for different time steps
• Figure 5: Context-aware grid map and CNN decision engine for AV decision making